This is an ambitious proposal for a 3 year multicenter study that will involve at least 12 sites. Using in vivo 1H-MRS, this consortium will explore hypotheses that relate various patterns of CNS cellular injury in individuals with HIV infection. The aims are: 1) To quantify regional levels of NAA, Cho, MI in the HIV-infected brain in a well defined cohort of subjects with ADC; 2) To correlated regional measurements of cellular metabolites with severity of neurocognitive impairment and systemic disease; 3) To examine the relationship of neurotoxicity, 1H-MRS results and neuropsychological testing; and 4) To determine the effects of memantine on the content of these performance. This consortium will study: (1) 80 subjects with mild or moderate dementia, enrolled in a randomized double-blind controlled trial of memantine act baseline and week 16; 2) 40 seropositive neurologically asymptomatic subjects with CD4 counts less than 200, and 3) 40 age-matched seronegative controls. Proton spectra will be obtained using a short echo STEAM sequence from the parietal cortex, the frontal centrum semiovale, and the basal ganglia. Metabolites from these regions will be quantified as ratios and absolute measurements and correlated with the severity of ADC and systemic disease, CSF/peripheral studies of surrogate markers and viral burden, and response to memantine and background antiretroviral therapy.